Fix Neo4j fulltext index name mismatch between creation and query

[Copilot]

The fulltext index is created with workspace-specific naming (entity_id_fulltext_idx_{workspace_label}) but search_labels queries the hardcoded name (entity_id_fulltext_idx). This causes IndexNotFound on new workspaces, forcing fallback to slow CONTAINS-based search.

Changes

• search_labels method (line 1644): Use workspace-specific index name matching creation logic
• Legacy migration (lines 243-270): Auto-migrate old entity_id_fulltext_idx indexes on init
• Early loop exit: Break when both current and legacy indexes found during init scan

Test Coverage

Added tests/test_neo4j_fulltext_index.py with integration tests for:

• Workspace-specific index creation and usage
• Chinese text search with CJK analyzer
• Multi-workspace index isolation
• CONTAINS fallback when index unavailable

# Before: hardcoded index name
index_name = "entity_id_fulltext_idx"

# After: workspace-scoped
index_name = f"entity_id_fulltext_idx_{workspace_label}"

Original prompt

check the commit to this branch

we might need to make chanegs:

After renaming the index, data migration must also be considered to ensure that indexes in legacy systems are properly updated to the new name.

lightrag/kg/neo4j_impl.py
):
"""Create a full-text index on the entity_id property with Chinese tokenizer support."""
index_name = "entity_id_fulltext_idx"
index_name = f"entity_id_fulltext_idx_{workspace_label}"
@chatgpt-codex-connector
chatgpt-codex-connector bot
6 hours ago
P1 Badge Full-text search uses stale index name

The full-text index is now created as entity_id_fulltext_idx_{workspace_label}, but search_labels still calls db.index.fulltext.queryNodes with the hard-coded name entity_id_fulltext_idx (see around line 1644), so on any newly created workspace where only the workspace-specific index exists, queries will throw IndexNotFound and always fall back to the slower CONTAINS-based path. The new index is never used, negating the change and degrading search performance for new databases.

Useful? React with 👍 / 👎.

is all of this true? can we add unit tests?

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.